IOP PUBLISHING JOURNAL OF PHYSICS: CONDENSED MATTER

J. Phys.: Condens. Matter 20 (2008) 362201 (5pp) doi:10.1088/0953-8984/20/36/362201

FAST TRACK COMMUNICATION

Imaging the atomic structure ofactivated carbonPeter J F Harris1, Zheng Liu2 and Kazu Suenaga2

1 Centre for Advanced Microscopy, J J Thomson Physical Laboratory, University of Reading,Whiteknights, Reading RG6 6AF, UK2 Research Centre for Advanced Carbon Materials, National Institute of Advanced IndustrialScience and Technology (AIST), Tsukuba 305-8565, Japan

E-mail: p.j.f.harris@reading.ac.uk and suenaga-kazu@aist.go.jp

Received 27 June 2008, in final form 30 July 2008Published 12 August 2008Online at stacks.iop.org/JPhysCM/20/362201

AbstractThe precise atomic structure of activated carbon is unknown, despite its huge commercialimportance in the purification of air and water. Diffraction methods have been extensivelyapplied to the study of microporous carbons, but cannot provide an unequivocal identification oftheir structure. Here we show that the structure of a commercial activated carbon can be imageddirectly using aberration-corrected transmission electron microscopy. Images are presentedboth of the as-produced carbon and of the carbon following heat treatment at 2000 ◦C. In the2000 ◦C carbon clear evidence is found for the presence of pentagonal rings, suggesting that thecarbons have a fullerene-related structure. Such a structure would help to explain the propertiesof activated carbon, and would also have important implications for the modelling of adsorptionon microporous carbons.

1. Introduction

Activated carbon is used on an enormous scale in gas andwater purification, metal extraction, medicine and many otherapplications [1]. It is prepared from a variety of carbonaceousprecursors, including coal, peat and nutshells which arecarbonized and then ‘activated’, either by oxidization with CO2

or steam, or by treatment with acids, bases or other chemicals.The resulting carbon can have a surface area of 1500 m2 g−1

or more, explaining its huge adsorptive capacity. Its preciseatomic structure, however, is unknown.

The first serious attempt to understand the structure ofcarbons produced by the pyrolysis of organic materials wasmade by Rosalind Franklin in the 1950s [2]. She showedthat these carbons fall into two distinct classes, which shecalled graphitizing and non-graphitizing. The kind of carbonfrom which activated carbon is derived is non-graphitizing,meaning that it cannot be transformed into crystalline graphiteeven at temperatures of 3000 ◦C and above. Franklin putforward a simple model of non-graphitizing carbon based onsmall graphitic crystallites joined together by cross-links, butdid not explain the nature of these cross-links. A later idea

was that the cross-links might consist of domains containingsp3-bonded atoms [3, 4], but this had to be discountedwhen neutron diffraction studies showed that non-graphitizingcarbons consist entirely of sp2 atoms [5]. A much morerecent suggestion [6–9] is that non-graphitizing carbon has astructure related to that of the fullerenes, in other words thatit consists of curved fragments containing pentagons and othernon-hexagonal rings in addition to hexagons, as illustrated infigure 1. Such a structure would explain the microporosityof the carbon, and many of its other properties. However,obtaining direct experimental support for this hypothesis isextraordinarily difficult. Both x-ray and neutron diffractionhave been extensively applied to non-graphitizing carbons, andin some studies the diffraction data has been interpreted interms of a structure containing non-hexagonal rings [10, 11].But definitive proof that the atoms are bonded in pentagonal orhexagonal rings cannot be obtained using diffraction methods.

Until relatively recently, imaging the atomic structure ofcarbons using transmission electron microscopy would alsohave been impossible, as the highest achievable resolutionwas greater than the C–C bond length (0.142 nm). However,the development of aberration correctors has improved the

0953-8984/08/362201+05$30.00 © 2008 IOP Publishing Ltd Printed in the UK1

J. Phys.: Condens. Matter 20 (2008) 362201 Fast Track Communication

Figure 1. Illustration of curved carbon fragments, containing pentagonal and heptagonal rings as well as hexagons.

resolution of TEM to the point where the direct imaging ofcarbon networks becomes possible. Ideally, such imagingshould be carried out at a lower accelerating voltage thanis normally used for high resolution TEM in order to avoidbeam damage. In previous studies, a TEM with a post-specimen aberration corrector has been operated at 120 kVto image the atomic structure of topological defects incarbon nanotubes [12] and of individual molecules insidenanotubes [13, 14]. Here we apply this technique, for the firsttime, to a conventional carbon. The carbon we have chosento study is the commercial activated carbon Norit GSX. Wedeliberately chose an activated carbon for study rather thanan untreated non-graphitizing carbon because the activationprocess results in an ‘open’ structure which is more amenableto imaging by HRTEM. Norit GSX is derived from peat, whichis carbonized in an inert atmosphere at 500 ◦C. Activationinvolves treating with steam at approximately 1000 ◦C andwashing in HCl. This produces a carbon with a surface area of950 m2 g−1 and an average pore size of approximately 3 nm.In addition to the as-received carbon, samples heated in Ar to2000 ◦C were imaged.

2. Experimental details

Specimens were prepared for TEM by dispersing the carbonin iso-propyl alcohol, mixing ultrasonically and depositingonto holey carbon support films. Images showing the overallmorphology of the carbons were recorded using the ReadingUniversity JEOL 2010 microscope, operated at 200 kV.Atomic resolution imaging was carried out using the TsukubaJEOL 2010F instrument, operated at 120 kV. This has apost-specimen aberration corrector (CEOS), giving a pointresolution of better than 0.14 nm at 120 kV. The Cs wasset to a value in the range from 0.5 to 5 μm in this work.Images were digitally recorded using a Gatan 894 CCD cameraunder a slightly under-focus condition (� f = −1 to −6 nm).Although an accelerating voltage of 120 kV is larger than theknock-on damage threshold for carbon materials, the beamdensity during the observation was suppressed to the lowestpossible level, to reduce damage to a minimum. We estimatethat the beam density during the observations was less than62 800 electrons nm−2 s−1.

3. Results

Micrographs of typical areas of the as-received and 2000 ◦Ccarbons are shown in figures 2(a) and (b) respectively. Theas-received carbon has a disordered and porous microstructureconsisting mainly of tightly curled single carbon layers. In the2000 ◦C carbon, the structure is still disordered, but with largerpores, bounded by more perfect carbon layers.

Obtaining high quality atomic resolution images of thefresh carbon was extremely challenging, owing to its highlydisordered structure. Nevertheless, a number of such imageswere recorded, and a typical example is shown in figure 3(a).At the edge of this fragment the individual rings of carbonatoms are resolved: the bright spots represent the centres ofthe rings. These bright spots do not form extended hexagonalarrays, which is consistent with the low crystallinity of thecarbon. However, some hexagonally arranged groups of spotscan be seen, as well as possible pentagonal arrangements(an example is arrowed). The images we have obtained ofthe fresh carbon are not yet of sufficient quality to justifycomparisons with simulations based on models of the structure.Nevertheless, our results demonstrate for the first time thatatomic level imaging of this kind of highly disordered carbonis possible.

Recording atomic resolution images of the 2000 ◦C-heatedcarbon was considerably easier than for the as-prepared carbon,due to the greater degree of crystallinity. As noted above, theeffect of heat treatment is to induce the growth of curved andfaceted carbon sheets, which enclose much larger pores thanin the fresh carbon. These sheets were frequently just a singlelayer thick, and in such cases images could be obtained whichshowed the true atomic structure. Two such images are shownin figures 3(b) and (c). In both images networks of hexagonalrings can clearly be seen. Quite often the rows of hexagonalrings observed in such images displayed some curvature; thiseffect can be seen in figure 3(c). This distortion indicates thatthe particle has a curved surface, rather than flat facets.

In some areas of the 2000 ◦C carbon there was clearevidence for the presence of five-membered rings. Figure 4(a)shows a small region in which at least two such ringsare apparently present. The area enlarged in figure 4(b)shows an arrangement of 5 bright spots surrounding a central

2

J. Phys.: Condens. Matter 20 (2008) 362201 Fast Track Communication

Figure 2. Conventional HRTEM micrographs showing general appearance of (a) fresh activated carbon, (b) activated carbon following heattreatment at 2000 ◦C.

Figure 3. (a) Aberration-corrected HRTEM micrograph of the fresh activated carbon, with the arrow indicating possible pentagonalarrangement of carbon rings. (b), (c) aberration-corrected micrographs of the 2000 ◦C-heated carbon.

spot. A good match was obtained with the simulated imagein figure 4(c), which was obtained from the structure infigure 4(d) using a standard multi-slice procedure. Here, thepentagon is oriented approximately parallel to the plane of theimage. A second area which contains a pentagonal structure is

shown in figure 4(e). In this case the central pentagonal ringis not visible, and we believe this is because the ring is tiltedaway from the plane of the image. Support for this comes fromthe reasonable match which can be seen between the image andthe simulated image in figure 4(f), obtained from the structure

3

J. Phys.: Condens. Matter 20 (2008) 362201 Fast Track Communication

a

b c d

e f

1 nm

g

Figure 4. (a) Aberration-corrected micrograph of a small area of the 2000 ◦C-heated carbon. (b) Enlarged region showing pentagonalarrangement of spots. (c) Simulated image of structure shown in (d). (e) Second region showing pentagonal arrangement. (f) Simulated imageof structure shown in (g).

in figure 4(g). We recognize that the agreement between theimages and simulations shown here is not exact, but this is tobe expected since we are not dealing with perfect crystallinestructures, but with disordered materials. Thus, the curvatureof the fragments of carbon will be affected by the morphologyof the surrounding material, and may not be identical to theidealized structures used for the image simulations. Taking thisinto account, we believe that the agreement is sufficiently goodto provide convincing evidence for the presence of pentagonalcarbon rings. It is difficult to make an accurate determinationof the ratio of pentagonal to hexagonal rings in the heat-treatedcarbon, as the number of micrographs recorded which clearly

show pentagons is rather small. On the basis of the imageswe do have, we would estimate that the ratio of pentagons tohexagons is approximately 1:50.

4. Conclusions

We have shown here, for the first time, that aberration-corrected TEM can be used to image the atomic structure ofa conventional carbon material, specifically the commercialactivated carbon Norit GSX. Atomic resolution images ofboth the as-produced carbon and of the carbon following heat

4

J. Phys.: Condens. Matter 20 (2008) 362201 Fast Track Communication

treatment at 2000 ◦C have been recorded. Images of the freshcarbon contained evidence of hexagonal rings, and possiblynon-hexagonal ones, but were difficult to interpret due tothe highly disordered structure. The 2000 ◦C-heated carboncontained larger hexagonal networks which were much morereadily imaged. There was also strong evidence in some ofthe images for the presence of pentagonal rings. The presenceof pentagons in the heat-treated carbon strongly suggests thatsuch rings were present in the original carbon. Once formed,pentagons tend to be kinetically ‘locked in’ to a hexagonallattice, since there is no easy mechanism for them to migrate.The effect of heat treatment on the fresh carbon is apparentlyto increase the size of the hexagonal networks, while leavingmany of the pentagons intact. Our results therefore support theidea, put forward more than ten years ago [6, 7] that this type ofcarbon has a fullerene-related structure. As already noted, sucha structure would help to explain many of the properties of non-graphitizing carbons, such as their microporosity, hardness andresistance to graphitization. A structure containing pentagonalrings would be naturally porous, due to the curvature of thecarbon layers, and would be relatively hard compared to othercarbons, owing to the absence of parallel graphene layers. Thestructure would also be resistant to graphitization, since non-hexagonal rings are extremely stable. The question arises asto the origin of the pentagonal rings. This is not known atpresent. In general, the process whereby organic materials aretransformed into carbon by heat treatment is not at all wellunderstood at the atomic level. Attempts have been made usingMonte Carlo methods to simulate the evolution of a polymerinto a microporous carbon structure containing non-hexagonalrings [15], and this kind of approach may eventually help toelucidate the mechanism.

The idea that microporous carbons have a fullerene-likestructure has important implications for the modelling ofadsorption on such carbons. Traditionally, such modellingexercises have utilized structural models derived from graphite,in which all the atoms are in hexagonal rings (e.g. [16]).The carbon pores are then assumed to have a slit-like shape,confined by parallel planes. If the fullerene-like modelsare correct, these ideas may have to be modified. Indeed,

theoretical studies have already been carried out which showthat a model structure containing fullerene-related elementsprovides a better basis for understanding adsorption onactivated carbon than the traditional models [17, 18].

Acknowledgments

This work is partially supported by CREST and Grant-in-Aid from MEXT(19054017). We thank Peter Campbell andMargriet Reimerink of Norit for helpful discussions.

References

[1] Marsh H and Rodrı́guez-Reinoso F 2006 Activated Carbon(Oxford: Elsevier)

[2] Franklin R E 1951 Proc. R. Soc. A 209 196[3] Ergun S and Tiensuu V H 1959 Acta Crystallogr. 12 1050[4] Ergun S and Alexander L E 1962 Nature 195 765[5] Mildner D F R and Carpenter J M 1982 J. Non-Cryst. Solids

47 391[6] Harris P J F and Tsang S C 1997 Phil. Mag. A 76 667[7] Harris P J F 1997 Int. Mater. Rev. 42 206[8] Harris P J F, Burian A and Duber S 2000 Phil. Mag. Lett.

80 381[9] Harris P J F 2005 Crit. Rev. Solid State Mater. Sci. 30 235

[10] Zetterstrom P, Urbonaite S, Lindberg F, Delaplane R G,Leis J and Svensson G 2005 J. Phys.: Condens. Matter17 3509

[11] Hawelek L, Koloczek J, Brodka A, Dore J C,Honkimaeki V and Burian A 2007 Phil. Mag. 87 4973

[12] Suenaga K, Wakabayashi H, Koshino M, Sato Y, Urita K andIijima S 2007 Nat. Nanotechnol. 2 358

[13] Sato Y, Suenaga K, Okubo S, Okazaki T and Iijima S 2007Nano Lett. 7 3704

[14] Liu Z, Yanagi K, Suenaga K, Kataura H and Iijima S 2007Nat. Nanotechnol. 2 422

[15] Kumar A, Lobo R F and Wagner N J 2005 Carbon 43 3099[16] Thomson K T and Gubbins K E 2000 Langmuir 16 5761[17] Terzyk A P, Furmaniak S, Gauden P A, Harris P J F,

Włoch J and Kowalczyk P 2007 J. Phys.: Condens. Matter19 406208

[18] Terzyk A P, Furmaniak S, Harris P J F, Gauden P A, Włoch J,Kowalczyk P and Rychlicki G 2007 Phys. Chem. Chem.Phys. 9 5919

5